Comparing a gradient boosting algorithm to the GOES FDC for wildfire detection
Authors: Asaf Vanunu, Boaz Nadler, Arnon Karnieli
Organizations: The Albert Katz International School for Desert Studies, The Jacob Blaustein Institutes for Desert Research, Sde Boker Campus, Ben-Gurion University of the Negev, Midreshet Ben-Gurion 84990, Israel · The Goldman Sonnenfeldt School of Sustainability and Climate Change, Ben-Gurion University of the Negev, Beer Sheva 8410501, Israel · The Remote Sensing Laboratory, French Associates Institute for Agriculture and Biotechnology of Drylands, The Jacob Blaustein Institutes for Desert Research, Sde Boker Campus, Ben-Gurion University of the Negev, 84990, Israel · Department of Computer Science and Applied Mathematics, Weizmann Institute of Science, Rehovot 7610001, Israel
Wildfires pose severe risks to human life, ecosystems, and property. This study presents a machine learning approach for wildfire detection from GOES ABI imagery. A CatBoost model was trained on a large dataset with thousands of ABI images and over 300,000 matching VIIRS fire detections. An evaluation on a separate dataset across five regions showed that the learned CatBoost model outperformed the operational GOES Fire Detection and Characterization (FDC) product. It achieved higher precision, recall, and F1 scores both within and outside the training area. The CatBoost model achieved F1 scores that were 0.16 to 0.38 higher than the GOES FDC in all regions. In addition, out of 51 historical fire events, the CatBoost detected 26 fires before both VIIRS and GOES FDC, compared to only six earlier detections by the GOES FDC. Importantly, the CatBoost model achieved accurate wildfire detection also during nighttime, whereas the GOES FDC obtained very low recall values, around 0.03. This study demonstrates that machine learning models may offer significant improvements over existing geostationary fire products, including higher accuracy, fewer false alarms, and earlier detection.
Figures & tables
Figure 1: Study area maps. The training states are marked in black, evaluation states in red, and the states used for both are marked in light blue.
Region
Months (2024)
GOES ABI Number of images
VIIRS Number of detected fires
Georgia
February, March, April
185
8211
Texas
March, April, May
325
9005
California-Oregon
July, August, September
363
53,259
Arizona-Utah
September, October, November
285
5309
British Columbia
July, August, September
115
1523
Table 1: Evaluation dataset
Figure 2: Map of the locations of 51 fire events with their identification numbers. For example, fire number 19 is the Bobcat Fire in Florida that occurred on 29 May 2021.
Figure 3: Rasterizing VIIRS fire detections onto the GOES ABI grid.
Figure 4: Illustration of the fire labeling process. Left: VIIRS fire detection at pixel (i,j) in the rasterized VIIRS matrix V . Right: A 3×3 window is applied in the GOES MCMI matrix G . The pixel G(i′,j′) with the highest FI value is labeled as fire.
Figure 5: Illustration of accuracy evaluation of a GEO-based fire detection algorithm using a 3×3 window. (left) Red pixels are GEO-based fire detections; (center) Pink pixels are rasterized VIIRS fire detections; (right) FN, TP, and FP are false negative, true positive, and false positive, respectively.
Figure 6: Comparison of performance metrics in terms of recall, precision, and F1 between the CatBoost model and the GOES FDC product in five different evaluation regions.
Figure 7: Example of CatBoost model predictions on the background of GOES-16 ABI FI images in Texas (left) and Georgia (right). VIIRS 375 m fire detections are shown as green squares. CatBoost prediction pixels are color-coded in red, purple, and blue for TP, FP, and FN, respectively.
Region
Daytime images
Nighttime images
Georgia
101
84
Texas
153
172
California-Oregon
180
183
Arizona-Utah
135
150
British Columbia
38
77
Table 2: Number of daytime and nighttime GOES ABI images for all study regions.
Figure 8: Recall, precision, and F1 scores of the CatBoost model and GOES FDC across all study regions. Panel A (upper) nighttime and Panel B (lower) daytime.
Figure 9: Comparison of performance metrics between CatBoost models using different number of fire pixels for training in five different evaluation regions.
Figure 10: First detection time difference between CatBoost and GOES FDC across fire events. Panel A shows the analysis using only the processed and saturated FDC categories, and Panel B includes the high category. Fire events that were detected simultaneously by CatBoost and GOES FDC have a time delta label of "0".
Fire event
First fire detection
Time difference (minutes)
Beech
CatBoost
300
Bentley
480
Davis
445
Smokehouse Creek
245
Airport
VIIRS
No CatBoost detection
Blue Mountain
Table 3: Summary of fire events detected (1) Only by CatBoost and VIIRS; (2) Only by VIIRS.
Figure 11: Probability of a fire pixel detected by the CatBoost and GOES FDC as a function of the minimum number of VIIRS fire detections inside a GOES ABI pixel ( k ). Evaluation regions: (A) Georgia, (B) Texas, (C) California-Oregon, (D) Arizona-Utah and (E) British Columbia.
Geostationary satellite observations are important for wildfire detection and monitoring. The current study evaluates machine learning models for MTG FCI near-real-time fire detection in 1- and 2-km spatial configurations and compares them with threshold-based algorithms. The models were trained and evaluated using VIIRS fire reference data across diverse ecological regions in Europe, Africa, and the Middle East. The key results are that 1-km models significantly outperform both their 2-km variants and operational threshold products. The constructed 1-km models achieved F1 scores higher by up to 0.36 compared to baseline products. Importantly, the 1-km models detected small fires with higher probability compared to competing models. Finally, the models robustly detected fires up to 260 min earlier than baseline products. To support opensource applications, our trained models are publicly available.
We present a deployed system for on-orbit wildfire detection aboard a nine-satellite commercial thermal infrared constellation, operating under demanding joint constraints: sub-megabyte model footprint, sub-150 ms per-batch TensorRT FP16 inference on an NVIDIA Jetson Xavier NX, and an end-to-end alert pipeline targeting under 10 minutes from satellite overpass to fire event communication. The system operates on uncalibrated mid-wave infrared (MWIR) single-band imagery at 200 m ground sampling distance, where fires frequently appear as sub-pixel or single-pixel thermal anomalies under extreme class imbalance -- challenges not addressed by the contextual thermal-thresholding pipelines (MODIS, VIIRS) that currently dominate operational fire monitoring. We present an empirical study of lightweight dense representation learning for this regime using a proprietary nine-satellite MWIR dataset. We compare dense masked autoencoding (DenseMAE) and a hybrid DenseMAE+EMA (exponential moving average) distillation variant, and evaluate representations via linear probing and full-distribution pixel-level average precision (AP) under extreme class imbalance. DenseMAE pretraining enables compact downstream models on the latency-accuracy Pareto frontier: our fastest SSL-pretrained model achieves 0.640 test AP and 0.69 event-level Fire-F1 with 65.34 ms latency per batch and a 0.52 MB engine, without pruning or compression. The best configuration reaches 0.699 AP and 0.744 Fire-F1 below 1 MB, outperforming a supervised baseline (0.650 AP) under comparable constraints.
Matthias Rötzer, Veronika Pörtge, Martin Ickerott +6
Over the past decades, the frequency of global wildfires has been increasing steadily. Therefore, if the fire can be detected and precisely located at an early stage, the potential hazards caused by it can be minimized to the greatest extent. The machine learning methods based on satellite images, due to their ability to automatically monitor extremely remote and vast areas, have shown great potential for application in the field of wildfire detection. To address this challenge, we proposed a new model named spectral-morphological attention U-Net(SMA-UNet), which includes a spectral attention module, a residual attention UNet backbone, a channel-spatial modulator, and a pair of differentiable morphological gates. We trained and evaluated this model with two datasets. These modules, excluding the backbone, are used to detect active fire events for the first time, especially the pair of differentiable morphological gates, which is innovatively developed. The proposed model achieved the highest scores in both datasets (e.g., intersection over union 75.16% in TS-SatFire, 22.50% in Sen2Fire). By conducting ablation studies of each module, we compared their independent contributions and tested their combinations. Ultimately, the integration of these modules yields a highly robust framework that significantly improves segmentation consistency across diverse and complex environmental conditions. Future work will focus on validating the proposed architecture across large-scale, multi-regional datasets from different satellite sensors to establish its broader generalizability for global wildfire detection.
Yugong Zeng, Jonathan Wu
Department of Electrical and Computer Engineering University of Windsor 401 Sunset Avenue Ontario, Canada N9B 3P4